Abstract
This article describes a Vocal Tract Length Nor malization (VTLN) procedure through frequency warping based on pitch estimates. This procedure aims to reduce the inter-speaker variability of speech signals in order to obtain a robust automatic speech recognition system. Two additional methods are also described: one for reducing the environment variability and another for compensating the coarticulation effects on connected word pronunciation. En vironment variability is compensated by explicitly modeling some frequent noise phenomena. Coarticulation phenomena compensation reduces speech signal variability by modeling events that result from coarticulation between adjacent mod els. Inter-speaker variability removal is performed by a traditional speaker normalization method, which consists in expanding or compressing the Mel filterbank bandwidths, in order to normalize the Vocal Tract Length (VTL) of each speaker. Most of the existing methods for VTL estimation are based on formant estimation, but the difficulty of formant estimation is a known performance limitation. The proposed method over comes such a problem since it estimates the warping factor through pitch. The recognition results, obtained for a tele phone digit recognition task (with phones and sub words as units), prove that this procedure leads to similar improve ments to those obtained with traditional methods based on formant estimates, actually outperforming them in some sit uations.
Highlights
The task of efficiently recognize spoken credit card num bers, telephone numbers or any other identification number is extremely important and requires an almost ideal recognition rate
Despite the good performance of the presently available Automatic Speech Recog nition (ASR) systems, the recognition robustness which allow the system to operate with an unlimited vocabulary, has not been fully achieved yet; if speaker independency and en vironment changes are taken into account
This work looks out for a valid contribution for solving the cen tral problem of ASR system robustness. This is achieved by modulating some variability agents present on speech signals, namely inter-speaker variability, environment variability and the variability imposed by an overwhelming number of coar
Summary
The task of efficiently recognize spoken credit card num bers, telephone numbers or any other identification number is extremely important and requires an almost ideal recognition rate. This work looks out for a valid contribution for solving the cen tral problem of ASR system robustness This is achieved by modulating some variability agents present on speech signals, namely inter-speaker variability, environment variability and the variability imposed by an overwhelming number of coar. Speaker normaliza tion aims to reduce the speech variability resulting from dif ferences between speakers - the so-called inter-speaker vari ability These differences are essentially related to vocal and nasal tract shape and length, vocal chord physiology and to gender and age. In this work it is investi gated the advantage of separating the models by speaker gen der.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.